Exclusion of a Transmembrane-Type Peptide from Ordered-Lipid

Quenching is shown in the two-phase range of compositions for Kp values that are near-infinite and in which the smaller domains have (solid line) infi...
0 downloads 0 Views 232KB Size
12376

Biochemistry 2003, 42, 12376-12390

Exclusion of a Transmembrane-Type Peptide from Ordered-Lipid Domains (Rafts) Detected by Fluorescence Quenching: Extension of Quenching Analysis to Account for the Effects of Domain Size and Domain Boundaries† Michael E. Fastenberg,‡ Hidehiko Shogomori,‡ Xiaolian Xu,‡ Deborah A. Brown,‡ and Erwin London*,‡,§ Department of Biochemistry and Cell Biology and Department of Chemistry, State UniVersity of New York at Stony Brook, Stony Brook, New York 11794-5215 ReceiVed May 5, 2003; ReVised Manuscript ReceiVed August 24, 2003

ABSTRACT: Sphingolipid/cholesterol-rich rafts are membrane domains thought to exist in the liquid-ordered state. To understand the rules governing the association of proteins with rafts, the behavior of a model membrane-inserted hydrophobic polypeptide (LW peptide, acetyl-K2W2L8AL8W2K2-amide) was examined. The distribution of LW peptide between coexisting ordered and disordered lipid domains was probed by measuring the amount of LW Trp fluorescence quenched by a nitroxide-labeled phospholipid that concentrated in disordered lipid domains. Strong quenching of the Trp fluorescence (relative to quenching in model membranes lacking domains) showed that LW peptide was concentrated in quencher-rich disordered domains and was largely excluded from ordered domains. Exclusion of LW peptide from the ordered domains was observed both in the absence and in the presence of 25-33 mol % cholesterol, indicating that the peptide is relatively excluded both from gel-state domains (which form in the absence of cholesterol) and from liquid-ordered-state domains (which form at high cholesterol concentrations). Because exclusion was also observed when ordered domains contained sphingomyelin in place of DPPC, or ergosterol in place of cholesterol, it appeared that this behavior was not strongly dependent on lipid structure. In both the absence and the presence of 25 mol % cholesterol, exclusion was also not strongly dependent upon the fraction of the bilayer in the form of ordered domains. To evaluate LW peptide behavior in more detail, an analysis of the effects of domain size and edges upon quenching was formulated. This analysis showed that quenching can be affected both by domain size and by whether a fluorescent molecule localized at domain edges. Its application to the quenching of LW peptide indicated that the peptide did not preferentially reside at the boundaries between ordered and disordered domains.

In many eukaryotic cell membranes, it is now believed that domains rich in unsaturated phospholipids coexist with lipid rafts, liquid-ordered domains that are rich in both relatively saturated sphingolipids and sterols. The association of proteins with rafts has been implicated as essential for a variety of cellular processes (1-7). Thus, understanding the principles that govern how lipids and proteins associate with rafts is important for understanding raft function. Some of these principles have been defined for lipid-lipid interaction. In particular, tight packing between relatively saturated lipids and cholesterol is a key feature that stabilizes raft formation (8, 9). When these lipids pack tightly, they form the Triton X-100-insoluble liquid-ordered state (8, 10-12). Certain principles governing the association of proteins with rafts have also been elucidated. Cell-derived detergentresistant membranes, which are believed to arise from rafts, are especially enriched in proteins anchored to membranes by saturated acyl chains (13, 14). Strong association of such proteins with detergent-insoluble sphingolipid/cholesterol† Supported by NIH Grants GM 48596 (E.L.) and GM 47897 (D.A.B.). * Corresponding author. Phone: (631) 632-8564. Fax: (631) 6328575. E-mail: [email protected]. ‡ Department of Biochemistry and Cell Biology. § Department of Chemistry.

rich domains has also been demonstrated using model membranes (11, 15-17). These observations have been explained by the ability of tightly packed, ordered-lipid domains to accommodate saturated-lipid anchors (11). In contrast, covalently linked prenyl groups, which have structures that should prevent their packing tightly within ordered-lipid domains, are not likely to contribute to the association of proteins with rafts (14, 18, 19). In general, transmembrane (TM)1 proteins also would not be expected to have the ability to pack tightly with lipids in ordered domains and thus should be excluded from lipid rafts. A lack of detergent insolubility has been used to show that the TM protein bacteriorhodopsin is excluded from liquidordered domains in Vitro (15). In addition, Van Duyl et al. recently demonstrated by detergent solubility that a hydrophobic peptide with an alternating Leu-Ala sequence was 1 Abbreviations: DOPC, 1,2-dioleoyl-sn-glycero-3-phosphocholine; DPH, diphenylhexatriene; DPPC, 1,2-dipalmitoyl-sn-glycero-3-phosphocholine; LW peptide, acetyl-K2W2L8AL8W2K2-amide; MLV, multilamellar vesicles; NBD-PE, N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2dipalmitoyl-sn-glycero-3-phosphoethanolamine; PBS, phosphate buffered saline; 12SLPC, 1-palmitoyl-2-(12-doxyl)stearoyl-sn-glycero-3-phosphocholine; rhodamine-PE, N-(lissamine rhodamine B)-1,2-dipalmitoylsn-glycero-3-phosphoethanolamine; SM, sphingomyelin; TM, transmembrane.

10.1021/bi034718d CCC: $25.00 © 2003 American Chemical Society Published on Web 10/04/2003

Protein Exclusion from Rafts excluded from ordered domains regardless of the extent of any hydrophobic mismatch between bilayer width and polypeptide length (20). Spectroscopic approaches avoiding potential artifacts arising from detergent-based methods are also valuable for such studies. In bilayers lacking cholesterol, fluorescence quenching by nitroxide-labeled lipids has been used to show that several different polypeptides are excluded from orderedgel phases (21-23). In a notable energy transfer study, Polozova and Litman found that at high cholesterol levels rhodopsin is depleted from ordered domains and enriched in disordered domains containing a highly polyunsaturated lipid (having 22:6 acyl chains) (24). Nevertheless, some TM proteins are believed to associate with lipid rafts (25, 26). In certain cases, the binding of such proteins to gangliosides, which are believed to be raftassociated, may result in raft association (27, 28). In other cases, covalent linkage of TM proteins to saturated acyl chains is necessary for raft association (14, 29-35). However, attachment to saturated acyl chains may not be the only factor in the raft association of such proteins. For example, caveolin-1 associates with ordered domains, yet mutations that prevent the attachment of saturated acyl chains to it do not prevent its association with rafts (36). In addition, changes in the amino acid sequence of TM segments in influenza hemagglutinin can abolish raft association without altering palmitoylation (37), and a quenching study suggested that a hydrophobic mismatch may affect the partition of gramicidin A′ between gel and fluid phases (38). In this paper, we investigated the behavior of a simple 25-residue hydrophobic peptide (LW peptide) to explore the interaction, or lack of interaction, of membrane proteins with rafts. To provide a strong fluorescence signal, LW peptide was designed with two Trp at the N-terminal boundary plus two Trp at the C-terminal boundary of its hydrophobic sequence. LW peptide is a member of the well-studied polyLeu class of peptides. This class of polypeptides forms stable TM helices (39-43). The distribution of LW peptide between domains was determined by measuring its intrinsic Trp fluorescence when it was inserted into model membranes containing both disordered domains that were enriched in a fluorescence quenching lipid and quencher-depleted ordered domains. These experiments revealed that LW peptide was relatively excluded from ordered domains under a variety of conditions. Lipid domain size can be drastically dependent on the exact experimental conditions (44), and rafts may be extremely small in cells (45, 46). Therefore, we also analyzed the impact of domain/raft size on quenching. This analysis showed that the quenching method should both be able to detect very small rafts and be able to determine whether molecules concentrate at raft boundaries. Our study indicates that fluorescence quenching should be a useful way to systematically investigate the relationship between the sequence/structure of TM proteins and the nature of their association with lipid rafts. EXPERIMENTAL PROCEDURES Materials. Dipalmitoylphosphatidylcholine (DPPC), dioleoylphosphatidylcholine (DOPC), brain sphingomyelin, 1-palmitoyl-2-(12-doxyl)stearoylphosphatidylcholine (12SLPC),

Biochemistry, Vol. 42, No. 42, 2003 12377 and cholesterol were purchased from Avanti Polar Lipids (Alabaster, AL). Androstenol was purchased from Sigma Chemical (St. Louis, MO). Ergosterol was purchased from Fluka Chemical (Ronkonkoma, NY). NBD-PE and rhodaminePE were purchased from Molecular Probes (Eugene, OR) or Avanti Polar Lipids. LW peptide, acetyl-K2W2L8AL8W2K2amide, was purchased from Research Genetics, Inc. (now Invitrogen, Carlsbad, CA). It was purified by reverse-phase high-pressure liquid chromatography using an 2-propanol/ water gradient as described previously (47). A fraction (about 20%) of the purified peptide molecules contained a one Leu deletion. The interaction of this slightly shorter peptide with lipid is likely to be very similar to that of the full-length LW peptide. Diphenylhexatriene (DPH) was purchased from Sigma-Aldrich (St. Louis, MO). Lipid concentrations were determined by dry weight, and peptide concentrations were determined by spectrophotometry using a molar absorptivity value at 280 nm of 5560 M-1 cm-1 per Trp residue. Lipids, DPH, and peptide were dissolved in ethanol and stored at -20 °C until use. Preparation of Multilamellar Vesicles (MLV). Ethanolic solutions of lipids and peptide were allowed to warm to room temperature for 1 h prior to use. Samples were prepared by mixing appropriate volumes of lipids and peptide or DPH into 12 × 75 mm disposable glass tubes, using calibrated glass capillary pipets. After mixing, the samples were dried under N2, after which 20 µL of chloroform was added to each tube. Samples were then redried, first with N2 and then under high vacuum for 1 h. Samples containing DPH were covered to protect them from light during the drying processes. To form MLV, 800 µL of PBS (10 mM Na phosphate, 150 mM NaCl, pH 7.4), warmed to 50-55 °C, was added to each sample. They were then immediately placed into a water bath (at 50-55 °C), covered, and incubated for 10 min. The samples were transferred to a multi-tube vortexer (VWR Scientific, West Chester, PA), placed in a 50-55 °C chamber, and agitated vigorously for 2-4 min. The samples were removed, briefly individually vortexed while exposed to room temperature, and finally allowed to cool. In rare cases in which the lipids did not disperse well after vortexing, samples were sonicated for 5-10 s in a bath sonciator (model G112SP1T, Laboratory Supplies, Hicksville, NY). Prior to measurements, samples were incubated at room temperature for at least an additional 45 min. Samples containing DPH were covered to protect them from light during this period. Preparation of Ethanol Dilution Vesicles. Ethanol dilution vesicles were prepared similarly to MLV, except that after the initial drying with N2, 16 µL of ethanol was added to redissolve the samples. Next, 784 µL of PBS warmed to 5055 °C was added to each sample, and they were then placed in a 50-55 °C bath for 5 min. As above, the samples were briefly vortexed manually and then allowed to cool to room temperature prior to fluorescence measurements. Fluorescence Quenching Experiments. Vesicle samples for thermal scanning experiments contained 50 µM total lipid (phospholipids plus sterol) plus 0.5-2 mol % peptide and/ or 0.25 mol % DPH. Unless otherwise noted, these samples were prepared by the MLV procedure described previously. In most cases, DPPC- or SM-containing samples with quencher (F samples) contained a 2:1:1 (mol/mol) [DPPC

12378 Biochemistry, Vol. 42, No. 42, 2003 or SM]/DOPC/12SLPC mixture with or without 25 mol % sterol. Corresponding samples without quencher (Fo samples) contained 1:1 DPPC/DOPC with or without 25 mol % sterol. In other experiments, lipid compositions of 3:1 mol:mol DPPC/12SLPC and Fo samples of 3:1 DPPC/DOPC were used. For both the 2:1:1 and the 3:1 compositions, control samples, which contained quencher but which could not form domains, were also prepared. These contained a 3:1 mixture of DOPC/12SLPC with or without 25 mol % sterol. Corresponding control Fo samples contained DOPC with or without 25 mol % sterol. For each lipid composition, background samples were prepared identically, except without peptide or DPH. Samples containing peptide or DPH were generally prepared in quadruplicate. (In some cases, DPH samples were prepared in duplicate.) Fluorescence in each sample was measured at 23 °C to assess sample-tosample variation, and the quadruplicates were then pooled. The fluorescence in pooled samples was then measured at a series of increasing temperatures as described next. Pooling samples was found to significantly reduce the variability of fluorescence intensities observed relative to that when individually prepared samples were used. Prior to pooling the samples, the standard deviations of the fluorescence intensities for the four individual samples generally ranged between 5 and 10% of the average F or Fo values. In the unusual cases in which a larger variation was observed, the experiments were usually repeated with an additional set of pooled quadruplicates, and values for the two pooled sets were averaged. Background values were found to be largely temperature independent and so only monitored at the lowest (starting) temperatures. Background values for peptide experiments were subtracted from the sample fluorescence intensities before F/Fo values were calculated. In the case of DPH experiments, backgrounds had a negligible fluorescence intensity. For experiments in which DPPC concentration was varied, ethanol dilution vesicles were prepared as described previously. Samples contained 50 µM total lipid (phospholipids plus sterol) and 2 mol % LW peptide dispersed in PBS. The sample volume was 800 µL. Samples, with or without 25 mol % cholesterol, contained various mixtures of DPPC and 12SLPC or DOPC and 12SLPC. Corresponding Fo samples contained DOPC in place of the 12SLPC. For each lipid composition, background samples were prepared identically, except without peptide. Their fluorescence was subtracted from the sample fluorescence intensities before F/Fo values were calculated. Fluorescence in each sample was measured at 23 °C. Fluorescence Measurements. Fluorescence emission spectra were measured on a Spex 212 Fluorolog fluorimeter using 10-mm excitation path length, 4-mm emission path length semimicro quartz cuvettes. Peptide fluorescence was measured at an excitation wavelength of 280 nm and emission wavelength of 340 nm. DPH fluorescence was measured at an excitation wavelength of 359 nm and an emission wavelength of 427 nm. Excitation slits were set at 2.5 mm (4.5-nm band-pass) for peptide or 1.25 mm for DPH. Emission slits were set at 5.0 mm. Temperatures were adjusted using a variable temperature water bath connected to a sample holder through which the bath solution circulated. When desired, sample temperatures were monitored with an electronic thermometer using a narrow probe placed within

Fastenberg et al. the sample solution, and the fluorescence was read 2 min after the samples reached the desired temperature. Sucrose Gradient Centrifugation. Sucrose gradient centrifugation was used to analyze the behavior of vesicles prepared by ethanol dilution. One preparation of vesicles contained a 100 µM 1:1 DPPC/DOPC or 3:3:2 DPPC/DOPC/ cholesterol (mol/mol) mixture plus 5 mol % LW peptide and 0.25 mol % rhodamine-PE. It was dispersed in 2.5 or 10% (w/v) sucrose. A second preparation of vesicles lacking peptide contained 100 µM 1:1 DPPC/DOPC or 3:3:2 DPPC/ DOPC/cholesterol plus 0.5 mol % NBD-PE. It was also dispersed in 2.5 or 10% (w/v) sucrose. (In other experiments, similar samples with or without peptides were prepared with the DPPC replaced by an equivalent amount of DOPC.) After 400 µL of the peptide-containing and peptide-lacking preparations were mixed, the external sucrose concentration was increased to 20% (w/v) with solid sucrose. The samples were then loaded on a 100 µL cushion of 50% (w/v) sucrose and overlayed with 3.6 mL of a 10-0% (w/v) linear sucrose gradient. Samples were centrifuged for 17 h at 38 000 rpm in a Beckman ultracentrifuge using a SW 60 swinging bucket rotor. Fractions of 200 µL were removed sequentially from the bottom of the tube. Trp fluorescence in each fraction was measured at an excitation wavelength of 280 nm and emission wavelength of 340 nm, NBD fluorescence was measured at an excitation wavelength of 470 nm and emission wavelength of 530 nm, and rhodamine fluorescence was measured at an excitation wavelength of 565 nm and emission wavelength of 585 nm. Control vesicle samples with and without each fluorophore showed that there was no significant contribution of any fluorophore to the background fluorescence intensity for any other fluorophore. RESULTS Quenching Analysis of the Partitioning of a Fluorescent Molecule between Ordered and Disordered Domains in Lipid Bilayers. Both domain formation and polypeptide association with specific lipid domains were assayed using fluorescence quenching ((9, 10, 18, 21, 48, 49)). In this approach, the partitioning of a fluorescent molecule is probed by the use of lipid vesicles containing a mixture of ordered and disordered domains. The vesicles contain lipids such as DPPC or sphingomyelin (SM), both of which tend to form ordered domains at room temperature, and 12SLPC, a lipid that carries a fluorescence quenching nitroxide group and that prefers to form and/or partition into disordered domains (10). Because of 12SLPC-induced quenching, fluorophores residing within 12SLPC-rich disordered domains fluoresce weakly, whereas those in the ordered domains, which are 12SLPC-poor, fluoresce more strongly (21, 38, 50). Experimentally, both the presence of domains and the partitioning of a fluorescent molecule between ordered and disordered domains are assessed by comparing quenching in samples containing lipid domains to quenching in analogous control lipid mixtures that form homogeneous bilayers lacking domains. It has been previously shown that a molecule associating to a significant degree with 12SLPCdepleted ordered domains is quenched less, and thus exhibits stronger fluorescence, in samples containing a mixture of ordered and disordered domains than in a homogeneous bilayer containing the same amount of 12SLPC (9, 10, 49).

Protein Exclusion from Rafts

FIGURE 1: Quenching of DPH fluorescence in lipid mixtures containing or lacking ordered domains and various sterols. Quenchercontaining vesicles were composed of DPH and (circles) 2:1:1 DPPC/DOPC/12SLPC (mol/mol) or (triangles) 3:1 DOPC/12SLPC with (A) 25 mol % cholesterol, (B) no sterol, (C) 25 mol % ergosterol, or (D) 25 mol % androstenol. Samples contained MLV at a concentration of 50 µM lipid with 0.25 mol % DPH and were dispersed in PBS. Samples were heated, and at each temperature the fluorescence in quencher (12SLPC)-containing vesicles (F) and in quencher-free (Fo) vesicles, in which 12SLPC was replaced by DOPC, was measured. F/Fo values were calculated after subtraction of the fluorescence in background samples lacking the fluorophore.

In contrast, a molecule residing only in the 12SLPC-rich disordered domains is quenched more strongly in a sample containing a mixture of domains than in a homogeneous bilayer. A more detailed discussion has been presented elsewhere (9, 10, 18, 21, 48, 49). Domain Formation in DPPC-Containing Bilayers. Preliminary studies (not shown) indicated that the partitioning of LW peptide between ordered and disordered domains could be assessed with maximal sensitivity in a mixture of 2:1:1 (mol/mol) DPPC/DOPC/12SLPC. At room temperature, we found previously that this mixture forms bilayers containing nearly equal amounts of DPPC-rich ordered domains and 12SLPC/DOPC-rich disordered domains (10). However, we had not examined the behavior of this lipid composition under the wide range of conditions used in the studies described next. To do so, we evaluated quenching of DPH fluorescence. (DPH is useful for such experiments because it partitions about equally well between most types of ordered or disordered domains and so will not perturb domain formation (21, 51).) Fluorescence quenching was assessed from the parameter F/Fo, which is the ratio of fluorescence in samples containing quencher (12SLPC) to that in samples in which quencher is replaced by DOPC. A low F/Fo value corresponds to a high degree of quenching. Figure 1 illustrates the quenching of DPH fluorescence in multilamellar vesicles (MLV) formed from 2:1:1 DPPC/ DOPC/12SLPC mixtures with or without 25 mol % sterol. In each case, higher F/Fo values were observed in the DPPCcontaining mixtures (circles) than in control mixtures con-

Biochemistry, Vol. 42, No. 42, 2003 12379

FIGURE 2: Quenching of LW peptide fluorescence in lipid vesicles containing or lacking ordered domains and various sterols. Quenchercontaining vesicles were composed of LW peptide and (circles) 2:1:1 DPPC/DOPC/12SLPC (mol/mol) or (triangles) 3:1 DOPC/ 12SLPC with (A) 25 mol % cholesterol, (B) no sterol, (C) 25 mol % ergosterol, or (D) 25 mol % androstenol. Samples contained MLV at a concentration of 50 µM lipid with 2 mol % LW peptide and were dispersed in PBS. Experiments were performed as described in Figure 1.

taining DOPC and 12SLPC (triangles) at 23 °C. The control mixtures form bilayers in a homogeneous disordered fluid state over the entire temperature range used (9, 10). Thus, the weak quenching of DPH fluorescence in the DPPCcontaining samples relative to control samples confirmed that the former contained domains at low temperature. At higher temperatures, quenching in the DPPC-containing samples increased to levels equivalent to that in the control samples, showing that domains in the DPPC-containing samples disappeared. High-temperature abolishes domains due to the melting of ordered domains and the mixing of their lipids with those in the disordered domains (9). Combined with previous studies of similar lipid mixtures (9), these results showed that at low temperatures DPPC-containing samples contained DPPC-rich ordered domains in equilibrium with DOPC/12SLPC-rich disordered domains. The difference between quenching in the DPPC-containing and control mixtures illustrated in Figure 1 also showed that the thermal dependence of quenching was significantly affected by the presence of 25 mol % sterol. Relative to bilayers lacking sterol (Figure 1B), in the presence of cholesterol and ergosterol there was less quenching at 23 °C and a higher temperature for the transition from a domaincontaining (weaker quenching) to a domain-lacking (stronger quenching) state (Figure 1A,C). Overall, the degree of ordered domain formation at 23 °C and the thermal stability of ordered domains decreased in the order ergosterol >cholesterol > no sterol > androstenol. This ranking is in good agreement with that found previously in similar lipid mixtures containing 15 mol % sterol (9, 52). Partitioning BehaVior of LW Peptide between Ordered and Disordered Domains: Effect of Sterol. Next, quenching was used to study the behavior of LW peptide. Figure 2 shows

12380 Biochemistry, Vol. 42, No. 42, 2003 the temperature dependence of the fluorescence quenching of LW peptide incorporated into MLV having the same compositions as in the DPH studies described previously. In all cases, at 23 °C, F/Fo was lower in DPPC-containing domain-forming samples (circles) than in the homogeneous control samples that lacked DPPC (triangles). As the temperature was increased, quenching in the DPPC-containing samples decreased, finally reaching values close to that in the control samples at 60 °C. As in the case of the quenching of DPH, this change in quenching is due to the melting of the ordered domains in the DPPC-containing samples at high temperature, which induces a state of homogeneous lipid mixing similar to that in the control sample. In experiments in which DPPC-containing samples were cooled to room temperature, this change was found to be largely reversible (not shown). As expected, control samples exhibited at most a very weak dependence of quenching upon temperature. The stronger quenching of LW peptide in DPPC-containing bilayers relative to control samples indicated that the LW peptide partitioned favorably into disordered domains and was relatively depleted from the ordered domains. Previous studies indicate that the ordered domains in lipid mixtures containing cholesterol are cholesterol-rich and exist in the liquid-ordered state (10, 11). Ordered domains forming in the absence of cholesterol are in the gel state. Therefore, we conclude that the LW peptide was excluded from both liquidordered and gel-state domains. It is noteworthy that quenching of LW peptide in the absence or presence of various sterols largely mirrored the level of ordered-domain formation. This was shown by the observation that the differences between F/Fo in the DPPCcontaining (circles) and control samples (triangles) decreased in an order (ergosterol g cholesterol > no sterol g androstenol) similar to that observed for DPH quenching. This pattern suggests that LW peptide behavior was influenced by sterol effects on domain formation rather than specific sterol-peptide interactions. Despite the parallels between the quenching patterns for DPH and LW peptide, it was apparent that the curves describing the temperature dependence of quenching LW peptide showed more gradual changes in F/Fo than for DPH. This indicated that the thermal process sensed by LW quenching was more gradual in terms of its temperature dependence than that sensed by DPH (see next and Discussion). It was also noted that F/Fo values for LW peptide in homogeneous bilayers in the absence of sterol (Figure 2B, triangles) were lower than in homogeneous samples containing sterol (Figure 2A,C,D, triangles) under equivalent conditions. The stronger quenching in the absence of sterol probably resulted from the fact that in the absence of sterol, 12SLPC molecules made up a higher fraction of the total lipid and so had a higher concentration within the bilayer. Strong Association of LW Peptide with Disordered Domains Is ObserVed under a Range of Conditions. These findings were next expanded by evaluating the partitioning of LW peptide under a variety of conditions. Both domain formation (as detected by DPH fluorescence, Figure 3D,E) and the preferential association of LW peptide with disordered domains (Figure 3A,B) were maintained when the amount of cholesterol was increased to 33 mol % (Figure

Fastenberg et al. 3A,D) or in small unilamellar vesicles prepared by ethanol dilution (Figure 3B,E). Other experiments showed that LW peptide also associated with disordered domains when DPPC was replaced by sphingomyelin (SM), a natural component of lipid rafts with phase behavior similar to that of DPPC (10) (Figure 3C). It should be noted that a thermal dependence of the quenching of the LW peptide similar to that observed with 2 mol % peptide was observed in DPPCcontaining samples with 25 mol % cholesterol when the concentration of peptide in the bilayer was decreased to 0.5 mol % (data not shown). Since LW peptide preferentially associates with disordered domains, it might be expected to destabilize ordered-domain formation. However, the DPH quenching showed that ordered-domain formation in DPPC-containing mixtures in the presence of 2 mol % LW peptide (Figure 3F) was not markedly different than that in its absence (Figure 1A). The inability of LW peptide to affect domain formation may stem from the fact that these samples had a relatively small amount of peptide relative to lipid in disordered domains. When the amount of lipid in disordered domains is small or polypeptide concentration is high, polypeptides may have much more dramatic effects upon domain formation (see next) (24). Effect of the Fraction of Ordered Lipid on the Partition BehaVior of LW Peptide. The effect of varying the mol fractions of DPPC and 12SLPC upon the quenching of LW peptide fluorescence was examined because under favorable conditions a profile of fluorescence versus DPPC mol fraction can be used to calculate Kp, the partition coefficient (21). This parameter is equal to the ratio of LW peptide concentration in disordered domains to that in ordered domains. The difference between quenching in DPPC/ 12SLPC mixtures and DOPC/12SLPC control mixtures depends on the value of Kp (10, 21) and should be observed throughout the range of lipid compositions over which ordered and disordered domains coexist. Previous studies have shown that in DPPC/12SLPC mixtures at 23 °C, bilayers are fully in the ordered state above 80 mol % DPPC (i.e., below 20 mol % 12SLPC), fully in the disordered state below 20 mol % DPPC (i.e., above 80 mol % 12SLPC), and contain coexisting ordered and disordered domains between 20 and 80 mol % DPPC () 80-20 mol % 12SLPC). This is true both in the presence and in the absence of cholesterol (10). Therefore, Kp-dependent quenching is expected in samples containing between 20 and 80 mol % 12SLPC. Figure 4 shows the effect of the 12SLPC mol fraction upon quenching of LW peptide fluorescence in vesicles prepared by ethanol dilution. LW peptide exhibited stronger quenching in DPPC/12SLPC mixtures (Figure 4, closed squares) than in the corresponding homogeneous DOPC/12SLPC control mixtures (Figure 4, open squares) over the entire 20-80 mol % 12SLPC range. This was true both in the absence (Figure 4A) and in the presence (Figure 4B) of cholesterol. These results showed that LW peptide strongly preferred to partition into disordered domains over the entire range of domain coexistence (i.e., that its preference for disordered domains did not strongly depend on what fraction of the bilayer was in the form of disordered domains). Theoretical curves describing the dependence of quenching upon Kp, calculated as described previously (21), are illustrated in Figure 5A. Comparison of Figures 4A to 5A

Protein Exclusion from Rafts

Biochemistry, Vol. 42, No. 42, 2003 12381

FIGURE 3: Quenching of LW peptide and DPH fluorescence in lipid mixtures containing or lacking ordered domains under various conditions. (A) Quenching of LW peptide fluorescence in samples containing 33 mol % cholesterol. Quencher-containing vesicles were composed of 2 mol % LW peptide and (circles) 2:1:1:2 DPPC/DOPC/12SLPC/cholesterol or (triangles) 3:1:2 DOPC/12SLPC/cholesterol. (B) Quenching of LW peptide fluorescence in vesicles prepared by ethanol dilution. Quencher-containing vesicles were composed of 2 mol % LW peptide and (circles) 2:1:1:1.33 DPPC/DOPC/12SLPC/cholesterol or (triangles) 3:1:1.33 DOPC/12SLPC/cholesterol. (C) Quenching of LW peptide fluorescence in quencher-containing vesicles containing SM. Vesicles contained 2 mol % LW peptide and (circles) 2:1:1:1.33 SM/DOPC/ 12SLPC/cholesterol or (triangles) 3:1:1.33 DOPC/12SLPC/cholesterol. (D) Quenching of DPH fluorescence in samples containing 33 mol % cholesterol. Quencher-containing vesicles were composed of 0.25 mol % DPH and (circles) 2:1:1:2 DPPC/DOPC/12SLPC/cholesterol or (triangles) 3:1:2 DOPC/12SLPC/cholesterol. (E) Quenching of DPH fluorescence in vesicles prepared by ethanol dilution. Quenchercontaining vesicles were composed of 0.25 mol % DPH and (circles) 2:1:1:1.33 DPPC/DOPC/12SLPC/cholesterol or (triangles) 3:1:1.33 DOPC/12SLPC/cholesterol. (F) Quenching of DPH fluorescence in vesicles also containing LW peptide. Quencher-containing vesicles were composed of 0.25 mol % DPH and (circles) 2:1:1:1.33 DPPC/DOPC/12SLPC/cholesterol plus 2 mol % LW peptide or (triangles) 3:1:1.33 DOPC/12SLPC/cholesterol with 2 mol % LW peptide. Other experimental conditions and procedures are as in Figures 1 and 2.

suggests that the Kp for LW peptide is about 10-20. That is, there is a 10-20-fold higher concentration of LW peptide in disordered domains than in ordered gel domains. Comparison of Figure 4, panels B to A suggests that partition between ordered and disordered domains in the presence of cholesterol is similar to that in its absence. However, a more precise estimate of Kp in the presence of cholesterol is not possible because analysis of quenching behavior in a threelipid component mixture (DPPC, 12SLPC, cholesterol) system is more complex than in a two-lipid component mixture (DPPC, 12SLPC) system, and Kp cannot be quantitatively modeled in the same manner (18). It should be noted that LW peptide fluorescence intensity in the absence of quencher (Fo) tends to support the conclusion that LW peptide partitions strongly into the disordered domains. Figure 4C shows that Fo values in bilayers composed solely of DPPC with or without cholesterol were somewhat more than half of those composed of DOPC with or without cholesterol. (We did not explore the origin of this increase in intensity in the disordered domains, but a likely possibility is that, due to peptide insolubility in ordered domains, when disordered domains are absent the peptides aggregate laterally, and this induces Trp selfquenching (53). Self-quenching is likely to be abolished when the peptides are more dispersed in the disordered domains.)

If LW peptide remained in ordered domains to an appreciable extent, then it would be predicted that Fo would only have increased gradually as a fraction of the bilayer in the disordered state was increased. Instead, the values of Fo increased rapidly as the DOPC concentration increased, reaching a value close to that in 100% DOPC at about 40 mol % DOPC. This is the behavior expected if the LW peptide partitioned strongly into DOPC-rich fluid domains, so that the LW peptide moved into disordered domains as soon as the DOPC concentration was high enough for an appreciable number of disordered domains to form. Possible EVidence for Perturbation of Lipid Phase BehaVior by LW Peptide. Interestingly, both with and without cholesterol, strong quenching of LW peptide in DPPCcontaining bilayers was observed when bilayers contained only 10 mol % 12SLPC (Figure 4). Our previous data suggested that DPPC/12SLPC mixtures with 10 mol % 12SLPC formed a homogeneous ordered state (10). However, it was difficult to specify the exact lipid composition at which disordered domains began to form, and it was possible that a small amount of 12SLPC-rich disordered state domains were present when the bilayers contained 10 mol % 12SLPC. Thus, one explanation for the strong quenching of LW peptide in bilayers of this composition is that the LW peptide was concentrated in 12SLPC-rich domains. The somewhat

12382 Biochemistry, Vol. 42, No. 42, 2003

FIGURE 4: Quenching of LW peptide fluorescence as a function of lipid composition. Ethanol dilution vesicles were prepared with various amounts of 12SLPC and DPPC (closed squares) or 12SLPC and DOPC (open squares) (A) Vesicles without cholesterol. (B) Vesicles with 25 mol % cholesterol. Samples contained 2 mol % LW peptide and 50 µM total lipid dispersed in PBS. F/Fo was measured at room temperature as described in Figure 1. Fo samples contained DOPC in place of 12SLPC. The x-axis shows the concentration of 12SLPC as a mol fraction of the total phospholipid. (C) Fo values for DPPC-containing samples (those indicated by filled squares in panels A and B). (Triangles) Fo values for panel A; (circles) Fo values for panel B. The x-axis gives the fraction of DOPC as a mol fraction of the total phospholipid.

weaker quenching at 10 mol % 12SLPC relative to that at higher mol % 12SLPC might be due to the fact that the quencher-to-peptide ratio is so low that there is not enough quencher lipid to fully surround the LW peptide molecules. An alternative is that the LW peptide had such a strong propensity for association with disordered lipid domains that it nucleated the formation of 12SLPC-rich domains under conditions in which they would not exist in the absence of peptide. These domains might be as little as a 12SLPC-rich boundary layer of lipid molecules surrounding each LW peptide. Effect of Domain Size Upon Fluorescence Quenching: Small Domain Size Can PreVent Calculation of Precise Kp Values. We were concerned that our interpretation of partition behavior might be incorrect if quenching was to be influenced by factors other than Kp, such as domain size. Domain size is important because small domains have relatively large boundary zones or edges. Unlike fluorescent molecules in the core of a domain, the fluorescence of molecules at the edge of a domain would be affected by quenchers both inside and outside the domain. To define the influence of domain size on quenching, an analysis of the relationship between quenching and domain

Fastenberg et al.

FIGURE 5: Calculated dependence of quenching of the fluorescence of a transmembrane helical polypeptide in bilayers composed of DPPC and 12SLPC. The fraction quencher is the fraction of the phospholipid molecules that are 12SLPC. In all four panels, F/Fo in a homogeneous bilayer that does not form domains is shown by long dashes. The other curves show F/Fo calculated using the equations derived in the Appendix for cases in which ordered domains contain 20 mol % 12SLPC, and disordered domains contain 80 mol % 12SLPC. (A) Effect of Kp upon quenching over the region of two-phase coexistence (compositions in which domains exist) for (open triangles) Kp ) 5; (crosses) Kp ) 10; (closed triangles) Kp ) 20; and (squares) Kp ) 50. (B) Effect of domain size upon quenching. It is assumed that the least abundant domains are small and are surrounded by a more abundant domain that is continuous, so that a shift between domains that are small occurs when equal amounts of ordered and disordered domains are present. Thus, for the case shown, disordered domains are small at the fraction 12SLPC < 0.5, while ordered domains are small at the fraction 12SLPC > 0.5. Quenching is shown in the two-phase range of compositions for Kp values that are near-infinite and in which the smaller domains have (solid line) infinite size so no molecules are near domain edges; (squares) 20% of molecules near the domain edge (2x/w ) 0.2); (closed triangles) 50% of molecules near the domain edge (2x/w ) 0.5); and (open triangles) 100% of molecules near the domain edge (2x/w ) 1). (C) The effect of different combinations of Kp and domain size: (diamonds) quenching in the two-phase range of compositions for Kp ) 50 and 2x/w ) 0.5 and (triangles) quenching in the two-phase range of the composition region for Kp ) 20 and near-infinite domain size. (D) Effect of location of the fluorescent molecule for domains of near-infinite size. Quenching is shown for fluorescent molecules: (long dashes) in a homogeneous bilayer; (solid line) present only in disordered domains (i.e., when Kp ) infinity); (dashes and dots) present only in ordered domains (i.e., when Kp ) 0); and (short dashes) only associating with domain edges (i.e., Kedge ) infinity).

size was formulated (Appendix). The case in which the less abundant lipid phase forms small domains and the more abundant lipid phase forms a continuous phase that surrounds these small domains was analyzed. Figure 5B,C shows the predicted effects of domain size upon 12SLPC quenching for a fluorescent polypeptide. In Figure 5B, the case in which the fluorescent polypeptide is fully excluded from ordered

Protein Exclusion from Rafts domains (i.e., for which the Kp is infinite) is considered. In this case, the qualitative effect of reducing the domain size down to very small dimensions is to reduce quenching relative to that observed with larger domains. For nitroxides, the quenching interaction distance is about equal to the diameter of a single lipid (54-56). As a result, a difference between nitroxide quenching in the presence of small domains (squares) relative to that in infinite size domains (solid line) would only become apparent when the domains have a width of 60% DPPC (44). In the future, a side-by-side comparison of raft formation by short-range quenchers and energy transfer has the potential to allow the estimation of the size of submicroscopic rafts. Factors Influencing Quenching: Association of Molecules with Raft Edges. We have also shown that fluorescent molecules specifically associating with domain edges would exhibit unique quenching behavior. Quenching would not only reflect Kp but also the association constant with the domain boundary, which we named Kedge. Although experimental results indicate that LW peptide does not concentrate near domain edges, it may be possible to use quenching to identify domain edge-seeking molecules in the future. The determination of whether molecules accumulate at raft edges is of particular interest because of the potential biological significance of raft boundaries for membrane structure and function. Mitchell and Litman have proposed that certain lipids with one saturated chain and one highly unsaturated chain orient such that their saturated chains cluster together with cholesterol (70). Since many natural phospholipids tend to have one saturated and one unsaturated chain, it is conceivable that some types may tend to concentrate at the edges of ordered domains with a sphingolipid/cholesterol core in an analogous manner. Such edgepreferring lipids could have a large impact upon raft structure and function because they would promote the formation of domain boundaries. With a fixed amount of raft-forming lipids, the introduction of edge-preferring lipids would tend to increase the domain perimeter-to-area ratio, requiring a decrease in domain size linked to an increase in domain number or a decrease in the smoothness of raft edges (8). Some proteins may also accumulate at domain edges. Interactions between a protein with a strong affinity for rafts and one that is excluded from rafts might cause them to accumulate at domain boundaries. In addition, TM proteins covalently modified by saturated acyl chains, or bound to raft-associated lipids, might orient such that the TM polypeptide remained outside a raft while the palmitate chains or bound lipid were immersed within a raft. Such edge localization could be functionally important. It could increase local protein concentration in one dimension as well as control the interaction of membrane components with each other. Difference between Temperature Dependence of LW and DPH Quenching. As noted in the Results, the thermal transitions detected by changes in the fluorescence of the LW molecules tended to be broader than those detected by DPH fluorescence. We showed that this was unlikely to result from inhomogeneous incorporation of peptide into a subset of vesicles. Instead, this difference may stem in part from the different partitioning behavior of the two probes. Theoretical analysis showed that the thermal dependence of

Fastenberg et al. the quenching of molecules partitioning strongly into fluid domains, like the LW peptide, has a slightly more gradual temperature dependence than that of molecules, which like DPH, partition equally between fluid and ordered domains (calculations not shown). This difference might also reflect a local perturbation of lipid phase behavior induced by LW peptide. In this regard, it is noteworthy that at high temperature, LW quenching in samples containing DPPC plus cholesterol or ergosterol seemed to indicate the presence of residual domains as judged by stronger quenching than in control samples. This difference in quenching was not noticeable for DPH fluorescence. One possible explanation for this is that LW molecules locally altered the mixing behavior of lipids such that small 12SLPC-enriched lipid domains locally persisted specifically around the peptide molecules. On the other hand, it is possible that the LW peptide was incorporated into a subset of vesicles enriched in 12SLPC. The observation that the strong quenching of peptide fluorescence was largely abolished at high temperatures in samples containing DPPC ruled out the possibility that LW peptide was incorporated into a subset of vesicles greatly enriched in 12SLPC relative to the average sample composition. However, incorporation of LW peptide into vesicles slightly enriched in 12SLPC could not be ruled out. Effect of Strong Partition into Fluid Domains on Domain Formation in Natural Membranes. The presence of TM proteins strongly excluded from ordered domains could significantly increase the formation of disordered domains in cell membranes. We saw possible evidence for such effects in bilayers with very high levels of saturated lipid (i.e., under conditions in which all, or almost all, of the bilayer would form ordered domains in the absence of peptide). The exact balance between ordered and disordered domains in vivo is unclear, but it has been proposed that a large fraction of the plasma membrane may consist of ordered domains (71-73). Therefore, it is possible that the high content of TM proteins in natural membranes influences lipid domain formation. Very recently, Ge et al. have suggested this as an explanation for electron spin resonance data obtained in mast cell membranes (74). APPENDIX: RELATIONSHIP BETWEEN DOMAIN SIZE AND QUENCHING To determine the effect of the domain size upon quenching, it is necessary to account for both the number of fluorophores in each domain and the number of fluorophores at the domain edges. The fraction of the fluorescent molecules in ordered and disordered domains is related to the total fraction of the bilayer that is in the ordered and disordered states and to the value of Kp for the fluorescent molecules. Kp is defined as the ratio of concentrations of the fluorescent molecule, which we name molecule M, in the disordered domains [Mdd] to that in ordered domains [Mod]. If the fraction of the bilayer that is in the disordered state is R, then that in the ordered state is 1 - R and

Kp ) [Mdd]/[Mod] ) (mol of Mdd/mol of Mod)(Aod/Add) ) (mol of Mdd/mol of Mod)((1 - R)/R) (1) where Add and Aod are the total areas of membrane that are in the form of disordered or ordered domains, respectively.

Protein Exclusion from Rafts

Biochemistry, Vol. 42, No. 42, 2003 12387

(For simplicity, we consider cases in which the quencher only affects fluorophores in the same leaflet of the bilayer. Under these conditions, the bilayer is essentially a twodimensional object, so that area is the analogue of volume in a three-dimensional solution.) Rearrangement of eq 1 allows the calculation of the fraction of fluorescent molecules located in ordered and disordered domains, fMod and fMdd, respectively.

fMod ) mol of Mod/(mol of Mod + mol of Mdd) ) (1 - R)/(1 - R + RKp) (2) fMdd ) mol of Mdd/(mol of Mod + mol of Mdd) ) RKp/(1 - R + RKp) (3) Two regions within a domain must be distinguished: the domain edge and the domain core. Considering the case in which a number of ordered domains exist and are surrounded by a single continuous disordered lipid domain, and in which molecule M has the same concentration in the core and at the edge of a domain, the fraction of fluorescent molecules within, but at the edges of, ordered domains (fMode) is

fMode ) fMod(Aode/Aod) ) ((1 - R)/(1 - R + RKp))(Aode/Aod) (4) where Aode/Aod is the ratio of the total area at the edge of ordered domains (Aode) to the total area of the ordered domains. The fraction of molecules in ordered domains, but not near their edge (i.e., in the core of ordered domains (fModc)), is then given by

fModc ) fMod - fMode ) fMod(1 - Aode/Aod) ) ((1 - R)/(1 - R + RKp))(1 - Aode/Aod) (5) Since the ratio of the total area in ordered domains to the total area in the disordered domains (Aod/Add) equals (1 R)/R, the fraction of molecules within disordered domains but at the edge of an ordered domain (fMdde) is then given by

fMdde ) fMdd(Adde/Add) ) fMdd(Adde/Aod)((1 - R)/R) ) (RKp/(1 - R + RKp))(Adde/Aod)((1 - R)/R) (6) where Adde is the total area at the edge of the disordered domains. Finally, the fraction of molecules in the disordered regions of the membranes that are not near the edge of an ordered domain (i.e., the fraction of molecules in the core of the disordered domain (fMddc)) is

fMddc ) fMdd - fMdde ) (RKp/(1 - R + RKp))(1 [(Adde/Aod)((1 - R)/R)]) (7) To make use of these equations, the fraction of a domain that is in its core and at the edge must be defined. The relative dimensions of the edge and core of a domain are related to the size of the domain and the range of the quenching interaction. For domains with the shape of long stripes of width w and a quenching interaction of range x (see Figure

FIGURE 8: Schematic illustration of bilayers containing stripeshaped domains. Lipids in disordered domains are shown as circles. Lipids in ordered domains are shown as squares. Lipids at domain edges are shown shaded. Labels: x ) width of domain edge, w ) width of domain, dd ) disordered domain, od ) ordered domain, dde ) edge of disordered domain, and ode ) edge of ordered domain. The local environment of four molecules (A-D) is also shown. Notice that molecule A in the core of the disordered domain is surrounded by six molecules in the disordered domain, and molecule B at the edge of the ordered domain is surrounded by two molecules in disordered domains and four molecules in ordered domains, etc.

9), the general relationship between domain size and edge size can be stated as

Aode/Aod ) 2x/w

(8)

Adde/Aod ) 2x/w

(9)

Notice that Aode ) Adde since the edge of the ordered and disordered domains have the same width (Figure 8). It is possible to define x and w in terms of lipid diameters (i.e., n lipid molecules thick or angstroms). Since natural phospholipids have a cross-sectional area close to 50-70 Å2 depending on the lipid state, the diameter of one lipid molecule is equivalent to 7-9 Å. To define the dimensions of x for the experiments in this paper, we must consider the case of quenching by nitroxidelabeled lipids. Nitroxide quenching is short-range, and roughly speaking only nitroxides attached to the nearest lipid neighbors of a fluorescent molecule should be close enough to quench fluorescence (54-56). Therefore, only fluorophores in the layer of the lipid at the immediate domain edge can be quenched by nitroxide-labeled lipids that are outside the domain in which they are located. In other words, the domain edge (x) is one lipid layer wide (see Figure 8). To estimate the effect of domain size upon fluorescence quenching, the quenching of molecules within the core and at the edges of both ordered and disordered domains must also be calculated. We consider a mixture of a nitroxidelabeled lipid and an unlabeled lipid and use a model in which the lipids in the bilayer exhibit typical close lateral packing, such that each lipid is surrounded by six others (Figure 8). The quenching of lipid-sized fluorescent molecules by

12388 Biochemistry, Vol. 42, No. 42, 2003

Fastenberg et al. Where F/Fo odc stands for F/Fo in the core of the ordered domains, etc. For long rectangular domains, substituting eqs 4-9 and the previous expressions describing the dependence of F/Fo upon C into eq 11 gives

F/Fo ) [((1 - R)/(1 - R + RKp))(1 - (2x/w))] × (1 - Co)6 + [((1 - R)/(1 - R + RKp))(2x/w)] × (1 - Co)4(1 - Cd)2 + [(RKp/(1 - R + RKp)) × (1 - (2x/w)((1 - R)/R))](1 - Cd)6 + [((1 - R)Kp/ (1 - R + RKp))(2x/w)](1 - Cd)4(1 - Co)2 (12)

FIGURE 9: Calculated dependence of quenching of a DPH-like probe (Kp ) 1) in bilayers composed of mixtures of DPPC and 12SLPC as a function of domain size. The fraction quencher is the fraction of phospholipid molecules that are 12SLPC. The dashed line shows quenching in a homogeneous bilayer that does not form domains. Quenching in the region of phase coexistence is calculated for the case in which ordered domains contain 20 mol % 12SLPC and disordered domains contain 80 mol % 12SLPC. It is assumed that the least abundant components (disordered domains or ordered domains) exist as small domains surrounded by a continuous domain formed by the more abundant component (i.e., for [fraction quencher] < 0.5, disordered domains are small, and for [fraction quencher] > 0.5, ordered domains are small). Quenching is shown in the two-phase (domain-forming) range of compositions in which domains (solid line) are of infinite size (i.e., with no molecules near domain edge); (open triangles) have 20% of molecules near the domain edge (2x/w ) 0.2); (crosses) have 50% of molecules near the domain edge (2x/w ) 0.5); or (closed triangles) have 100% of molecules near the domain edge (2x/w ) 1).

nitroxide-labeled lipids can be then approximated by the expression F/Fo ) (1 - C)6 where C is the local concentration of the quencher lipid in mol fraction units (22, 54). This expression is valid for a fluorescent molecule in the core of a domain (see Figure 8, molecules A and D) if it is fully quenched when any of its six neighbors carries a nitroxides (22, 54). At the edge of a domain, F/Fo depends on how many neighbors are within the domain, how many are outside the domain, and the value of C within and outside the domain. For domains with straight boundaries (e.g., a long rectangular stripe), four lipids will be within the same domain as the fluorophore, and two will be outside the domain (see Figure 8, molecules B and C). Thus, at the domain edge, F/Fo ) (1 - Cinside domain)4(1 - Coutside domain)2. Combining the previous equations allows the calculation of the dependence of quenching upon domain size. For simplicity, we assume that the fluorescent molecule in ordered and disordered domains has the same quantum yield when in the absence of a quencher. The general expression for F/Fo in a sample is then

F/Fo ) ∑ (fraction of fluorescent molecules in an environment) (F/Fo in that environment) ) ∑fMi(F/Fo)i (10) Where i represents each environment, and the sum is over all environments. This can be expanded into four terms

F/Fo ) fModc(F/Fo)odc + fMode(F/Fo)ode + fMddc(F/Fo)ddc + fMdde(F/Fo)dde (11)

where Co is the concentration of the quencher in the ordered domains (as a mol fraction of total lipid in the domain), and Cd is the concentration of the quencher in the disordered domains. Notice that this equation is not valid when 2x > w. If 2x > w, the area at the edge of a domain would exceed the total area within that domain, which is physically impossible. Analogous equations can be developed for the case in which a bilayer contains disordered domains surrounded by a continuous ordered domain. In this case, w is the width of the disordered domain. The fraction of fluorescent molecules within, but at the edge of the disordered domains, is given by

fMdde ) fMdd(Adde/Add) ) (RKp/(1 - R + RKp))(Adde/Add) (13) The fraction of molecules in the core of the disordered domains is given by

fMddc ) fMdd - fMdde ) fMdd(1 - Adde/Add) ) ((RKp)/(1 - R + RKp))(1 - Adde/Add) (14) The fraction of molecules in the ordered domains, but at the edge of a disordered domain, is given by

fMode ) fMod(Aode/Aod) ) fMod(Aode/Add)(R/(1 - R)) ) ((1 - R)/(1 - R + RKp))(Aode/Add)(R/(1 - R)) (15) Finally, the fraction of molecules in the core of the ordered regions is given by

fModc ) fMod - fMode ) ((1 - R)/(1 - R + RKp)) ((1 - R)/(1 - R + RKp))(Aode/Add)(R/(1 - R)) ) ((1 - R)/(1 - R + RKp))(1 - (Aode/Add)(R/(1 - R))) (16) Substitution of these expressions into eq 11 gives

F/Fo ) [((1 - R)/(1 - R + RKp))(1 - (R/(1 - R)) × (2x/w))](1 - Co)6 + [(R/(1 - R + RKp))(2x/w)] × (1 - Co)4(1 - Cd)2 + [((RKp)/(1 - R + RKp)) × (1 - 2x/w)](1 - Cd)6 + [(RKp/(1 - R + RKp))(2x/w)] × (1 - Co)2(1 - Cd)4 (17) Figure 9 shows the effect of domain size on the fluorescence quenching of a molecule that (like DPH) has a Kp close to 1. Notice that edge effects become significant when 2x/w

Protein Exclusion from Rafts

Biochemistry, Vol. 42, No. 42, 2003 12389

is greater than 0.2 (i.e., for domains that are 10 molecules wide) and that edge effects distort quenching strongly when 2x/w is as large as 0.5-1 (crosses and filled triangles). In fact, when domains are very small (filled triangles), the quenching behavior approaches that in homogeneous bilayers. Also notice that at a quencher mol fraction of 0.5, the small domain size can result in anomalous quenching, such that there is increased quenching relative to that in homogeneous bilayers under conditions in which the presence of large domains decreases the quenching relative to that in homogneous bilayers. The quenching equations derived previously must be modified for application to the Trp fluorescence of transmembrane polypeptides or proteins. As a consequence of their larger cross-sectional area relative to lipids, a lipidexposed Trp residue in a polypeptide will have fewer lipid neighbors than a lipid-sized molecule. Previous estimates give a crude value of 1-2 lipid neighbors per Trp for membrane proteins (54), and the data for the LW peptide used in this report also fits a value close to 2 (calculation not shown). Thus, for the Trp fluorescence of LW peptide in the core of a domain, F/Fo ) (1 - C)2. At the edge of a domain, where half of the lipid neighboring a Trp would be inside the domain and half outside, one can estimate that F/Fo ) (1 - Cinside domain)(1 - Coutside domain). In this case, eqs 12 and 17 become

F/Fo ) [((1 - R)/(1 - R + RKp))(1 - (2x/w))] × (1 - Co)2 + [(((1 - R)(1 + Kp))/(1 - R + RKp)) × (2x/w)](1 - Co)(1 - Cd) + [(RKp/(1 - R + RKp)) × (1 - (2x/w)((1 - R)/R))](1 - Cd)2 (18) and

F/Fo ) [((1 - R)/(1 - R + RKp))((1 - (R/(1 - R))) × (2x/w))](1 - Co)2 + [(R(1 + Kp)/(1 - R + RKp)) × (2x/w)](1 - Co)(1 - Cd) + [(RKp/(1 - R + RKp)) × (1 - 2x/w)](1 - Cd)2 (19) respectively. The previous analysis only considered the quenching of molecules whose behavior can be fully defined by their partition between ordered and disordered domains. A different fluorescence quenching pattern would be observed for a molecule is located at the boundaries between ordered and disordered domains. When domains are present, F/Fo for a polypeptide that sits exclusively at the domain edges would approximately be given by

F/Fo ) (1 - Co)(1 - Cd)

(20)

since in binary lipid mixtures, Co and Cd are invariant when ordered and disordered domains coexist, and F/Fo would be independent of the quencher concentration for these compositions. The analysis of the dependence of the quenching upon domain size can also be extended to circular domains. For circular ordered domains of radius r that are embedded in a

continuous disordered phase, the following equations apply:

Aode/Aod ) (πr2 - π(r - x)2)/πr2 ) 1 - (1 - x/r)2

(21)

Adde/Aod ) (π(r + x)2 - πr2)/πr2 ) (1 + x/r)2 - 1

(22)

The exact solution for the dependence of the quenching on the domain radius is algebraically complex for circular domains in a hexagonally packed lipid lattice. However, it is possible to crudely estimate how many lipid molecules must be in a circular domain to allow the domain to be easily detected by nitroxide lipid-induced quenching. On the basis of the preceding analysis, it can be estimated that to easily detect ordered domains surrounded by a continuous disordered domain, Aode/Aod < 0.5, for a probe having Kp ) 1. Solving eq 21 with Aode/Aod ) 0.5 gives x/r ) 0.293. If x is approximately 8 Å (see previously), then r is about 25-30 Å. This means that the area of a domain (πr2) must be about 2000 Å2 for easy detection. Since lipid molecules occupy about 70 Å2, this is equivalent to a domain of about 30 lipid molecules, showing that nitroxide quenching should be able to detect very small domains. Similarly, if edge effects are negligible when Aode/Aod < 0.2, then nitroxide quenching will not be significantly influenced by edges when r > 75 Å (i.e., for domains of over 250 lipid molecules). It should be noted that the analysis described in eqs 1022 ignores corrections needed when the quantum yield of a fluorophore is significantly different in the ordered and disordered state. The corrected form of eq 10 would be

F/Fo ) ∑(fMi(F/Fo)iNi)/∑(fMiNi)

(10a)

where Ni ) Foi/Fo1, which is the ratio of Fo values in environment i to that in the first environment. Use of the corrected form of eq 10 is unnecessary when the quantum yield difference is small or when the fluorescent molecules partition very strongly into one particular type of domain. Since Fo is proportional to the quantum yield, in practice a correction can be ignored when Fo is not dependent upon the fraction of the bilayer that is in the disordered state (R). Since Fo for the LW peptide was relatively constant over the entire range in which disordered and ordered domains coexist (i.e., 12SLPC concentrations between 0.2 and 0.8) (see Figure 4C), we did not attempt to calculate corrected Kp values for LW peptide. A second reason not to correct Kp was that the corrected previous equation assumes that the difference in quantum yield is not dependent upon the concentration of the fluorescence molecule in a particular domain. We suspect that the LW peptide aggregates within ordered bilayers and that this aggregation results in selfquenching that decreases Trp quantum yield. (Self-quenching of Trp at high local concentration has been observed previously (53).) If this is correct, and LW peptide concentration within ordered domains is greatly decreased in the presence of disordered domains (due to its partition out of the ordered domains), then the quantum yield correction would exaggerate how much of the LW peptide was in the ordered domains. REFERENCES 1. Brown, D. A., and London, E. (1998) Annu. ReV. Cell DeV. Biol. 14, 111-36.

12390 Biochemistry, Vol. 42, No. 42, 2003 2. Grassme, H., Jekle, A., Riehle, A., Schwarz, H., Berger, J., Sandhoff, K., Kolesnick, R., and Gulbins, E. (2001) J. Biol. Chem. 276, 20589-96. 3. Grassme, H., Jendrossek, V., Riehle, A., von Kurthy, G., Berger, J., Schwarz, H., Weller, M., Kolesnick, R., and Gulbins, E. (2003) Nat. Med. (N. Y.) 9, 322-30. 4. Simons, K., and Toomre, D. (2000) Nat. ReV. Mol. Cell Biol. 1, 31-9. 5. Niyogi, K., and Hildreth, J. E. (2001) J. Virol. 75, 7351-61. 6. Drevot, P., Langlet, C., Guo, X. J., Bernard, A. M., Colard, O., Chauvin, J. P., Lasserre, R., and He, H. T. (2002) EMBO J. 21, 1899-908. 7. Gomez-Mouton, C., Abad, J. L., Mira, E., Lacalle, R. A., Gallardo, E., Jimenez-Baranda, S., Illa, I., Bernad, A., Manes, S., and Martinez, A. C. (2001) Proc. Natl. Acad. Sci. U.S.A. 98, 96427. 8. Brown, D. A., and London, E. (1998) J. Membr. Biol. 164, 10314. 9. Xu, X., and London, E. (2000) Biochemistry 39, 843-9. 10. Ahmed, S. N., Brown, D. A., and London, E. (1997) Biochemistry 36, 10944-53. 11. Schroeder, R., London, E., and Brown, D. (1994) Proc. Natl. Acad. Sci. U.S.A. 91, 12130-4. 12. London, E., and Brown, D. A. (2000) Biochim. Biophys. Acta 1508, 182-95. 13. Brown, D. A., and Rose, J. K. (1992) Cell 68, 533-44. 14. Melkonian, K. A., Ostermeyer, A. G., Chen, J. Z., Roth, M. G., and Brown, D. A. (1999) J. Biol. Chem. 274, 3910-7. 15. Schroeder, R. J., Ahmed, S. N., Zhu, Y., London, E., and Brown, D. A. (1998) J. Biol. Chem. 273, 1150-7. 16. Wang, T. Y., Leventis, R., and Silvius, J. R. (2001) Biochemistry 40, 13031-40. 17. Dietrich, C., Volovyk, Z. N., Levi, M., Thompson, N. L., and Jacobson, K. (2001) Proc. Natl. Acad. Sci. U.S.A. 98, 10642-7. 18. Wang, T. Y., Leventis, R., and Silvius, J. R. (2000) Biophys. J. 79, 919-33. 19. Moffett, S., Brown, D. A., and Linder, M. E. (2000) J. Biol. Chem. 275, 2191-8. 20. van Duyl, B. Y., Rijkers, D. T., de Kruijff, B., and Killian, J. A. (2002) FEBS Lett. 523, 79-84. 21. London, E., and Feigenson, G. W. (1981) Biochim. Biophys. Acta 649, 89-97. 22. London, E., and Feigenson, G. W. (1981) Biochemistry 20, 193948. 23. Florine, K. I., and Feigenson, G. W. (1987) Biochemistry 26, 2978-83. 24. Polozova, A., and Litman, B. J. (2000) Biophys. J. 79, 2632-43. 25. Langlet, C., Bernard, A. M., Drevot, P., and He, H. T. (2000) Curr. Opin. Immunol. 12, 250-5. 26. Simons, K., and Ikonen, E. (1997) Nature 387, 569-72. 27. Miljan, E. A., Meuillet, E. J., Mania-Farnell, B., George, D., Yamamoto, H., Simon, H. G., and Bremer, E. G. (2002) J. Biol. Chem. 277, 10108-13. 28. Ghislain, J., Lingwood, C. A., and Fish, E. N. (1994) J. Immunol. 153, 3655-63. 29. Rousso, I., Mixon, M. B., Chen, B. K., and Kim, P. S. (2000) Proc. Natl. Acad. Sci. U.S.A. 97, 13523-5. 30. Zhang, W., Trible, R. P., and Samelson, L. E. (1998) Immunity 9, 239-46. 31. Niethammer, P., Delling, M., Sytnyk, V., Dityatev, A., Fukami, K., and Schachner, M. (2002) J. Cell Biol. 157, 521-32. 32. Li, M., Yang, C., Tong, S., Weidmann, A., and Compans, R. W. (2002) J. Virol. 76, 11845-52. 33. Fragoso, R., Ren, D., Zhang, X., Su, M. W., Burakoff, S. J., and Jin, Y. J. (2003) J. Immunol. 170, 913 - 21. 34. Arcaro, A., Gregoire, C., Bakker, T. R., Baldi, L., Jordan, M., Goffin, L., Boucheron, N., Wurm, F., van der Merwe, P. A., Malissen, B., and Luescher, I. F. (2001) J. Exp. Med. 194, 148595. 35. Shrimpton, C. N., Borthakur, G., Larrucea, S., Cruz, M. A., Dong, J. F., and Lopez, J. A. (2002) J. Exp. Med. 196, 1057-66.

Fastenberg et al. 36. Dietzen, D. J., Hastings, W. R., and Lublin, D. M. (1995) J. Biol. Chem. 270, 6838-42. 37. Scheiffele, P., Roth, M. G., and Simons, K. (1997) EMBO J. 16, 5501-8. 38. Dibble, A. R., Yeager, M. D., and Feigenson, G. W. (1993) Biochim. Biophys. Acta 1153, 155-62. 39. Huschilt, J. C., Millman, B. M., and Davis, J. H. (1989) Biochim. Biophys. Acta 979, 139-41. 40. Ren, J., Lew, S., Wang, J., and London, E. (1999) Biochemistry 38, 5905-12. 41. Ren, J., Lew, S., Wang, Z., and London, E. (1997) Biochemistry 36, 10213-20. 42. Webb, R. J., East, J. M., Sharma, R. P., and Lee, A. G. (1998) Biochemistry 37, 673-9. 43. Mall, S., Broadbridge, R., Sharma, R. P., Lee, A. G., and East, J. M. (2000) Biochemistry 39, 2071-8. 44. Feigenson, G. W., and Buboltz, J. T. (2001) Biophys. J. 80, 277588. 45. Edidin, M. (2001) Trends Cell Biol. 11, 492-496. 46. Jacobson, K., and Dietrich, C. (1999) Trends Cell Biol. 9, 8791. 47. Lew, S., and London, E. (1997) Anal. Biochem. 251, 113-6. 48. London, E., Brown, D. A., and Xu, X. (2000) Methods Enzymol. 312, 272-90. 49. London, E. (2002) Curr. Opin. Struct. Biol. 12, 480-6. 50. Feigenson, G. W. (1983) Biochemistry 22, 3106-12. 51. Lentz, B. R., Barenholz, Y., and Thompson, T. E. (1976) Biochemistry 15, 4521-8. 52. Xu, X., Bittman, R., Duportail, G., Heissler, D., Vilcheze, C., and London, E. (2001) J. Biol. Chem. 276, 33540-6. 53. London, E. (1986) Anal. Biochem. 154, 57-63. 54. London, E., and Feigenson, G. W. (1981) Biochemistry 20, 19328. 55. Yeager, M. D., and Feigenson, G. W. (1990) Biochemistry 29, 4380-92. 56. Chattopadhyay, A., and London, E. (1987) Biochemistry 26, 3945. 57. Mabrey, S., Mateo, P. L., and Sturtevant, J. M. (1978) Biochemistry 17, 2464-8. 58. Mall, S., Broadbridge, R., Sharma, R. P., East, J. M., and Lee, A. G. (2001) Biochemistry 40, 12379-86. 59. Florine-Casteel, K., and Feigenson, G. W. (1988) Biochim. Biophys. Acta 941, 102-6. 60. Huang, N. N., Florine-Casteel, K., Feigenson, G. W., and Spink, C. (1988) Biochim. Biophys. Acta 939, 124-30. 61. Spink, C. H., Yeager, M. D., and Feigenson, G. W. (1990) Biochim. Biophys. Acta 1023, 25-33. 62. Wang, T.-Y., and Silvius, J. R. (2003) Biophys. J. 84, 367-78. 63. Anderson, R. G., and Jacobson, K. (2002) Science 296, 1821-5. 64. Brown, D. A., and London, E. (2000) J. Biol. Chem. 275, 172214. 65. Milhiet, P. E., Giocondi, M. C., and Le Grimellec, C. (2002) J. Biol. Chem. 277, 875-8. 66. Rinia, H. A., and de Kruijff, B. (2001) FEBS Lett. 504, 194-9. 67. Yuan, C., and Johnston, L. J. (2000) Biophys. J. 79, 2768-81. 68. Yuan, C., and Johnston, L. J. (2001) Biophys. J. 81, 1059-69. 69. Silvius, J. R., del Giudice, D., and Lafleur, M. (1996) Biochemistry 35, 15198-208. 70. Mitchell, D. C., and Litman, B. J. (1998) Biophys. J. 75, 896908. 71. Gidwani, A., Holowka, D., and Baird, B. (2001) Biochemistry 40, 12422-9. 72. Hao, M., Mukherjee, S., and Maxfield, F. R. (2001) Proc. Natl. Acad. Sci. U.S.A. 98, 13072-7. 73. Rietveld, A., and Simons, K. (1998) Biochim. Biophys. Acta 1376, 467-79. 74. Ge, M., Gidwani, A., Brown, H. A., Holowka, D., Baird, B., and Freed, J. H. (2003) Biophys. J. 85, 1278-1288. BI034718D